Arizona Ex Rel. Henry Darwin v. Usepa

FOR PUBLICATION

UNITED STATES COURT OF APPEALS
FOR THE NINTH CIRCUIT

STATE OF ARIZONA, EX REL. HENRY No. 13-70366
R. DARWIN, Director, Arizona
Department of Environmental
Quality,
Petitioner,

v.

U.S. ENVIRONMENTAL PROTECTION
AGENCY; LISA P. JACKSON,
Administrator, United States
Environmental Protection Agency,
Respondents,

NATIONAL PARKS CONSERVATION
ASSOCIATION; SIERRA CLUB,
Respondent-Intervenor.
2 ARIZONA EX REL. DARWIN V. USEPA

SALT RIVER PROJECT No. 13-70410
AGRICULTURAL IMPROVEMENT AND
POWER DISTRICT, OPINION
Petitioner,

v.

U.S. ENVIRONMENTAL PROTECTION
AGENCY,
Respondent,

NATIONAL PARKS CONSERVATION
ASSOCIATION; SIERRA CLUB,
Respondent-Intervenors.

On Petition for Review of an Order of the
Environmental Protection Agency

Argued and Submitted
March 9, 2015—San Francisco, California

Filed February 24, 2016

Before: Marsha S. Berzon, Jay S. Bybee,
and John B. Owens, Circuit Judges.

Opinion by Judge Berzon
 ARIZONA EX REL. DARWIN V. USEPA 3

SUMMARY*

Environmental Law

The panel denied consolidated petitions for review of a
Final Rule, promulgated by the Environmental Protection
Agency under the Clean Air Act, that partially disapproved
Arizona’s regional haze State Implementation Plan (“SIP”) –
setting forth emission limits and other measures – and issued
a Federal Implementation Plan (“FIP”) in place of the
disapproved SIP elements.

The panel held that the EPA did not act arbitrarily and
capriciously when it disapproved in part the SIP’s “best
available retrofit technology” for the Coronado Generating
Station, a coal-fueled power plant located in Eastern Arizona,
and when it issued a replacement FIP as to the disapproved
portions. The panel also held that the EPA did not err
procedurally in promulgating the FIP in the same rule as its
partial disapproval of the SIP.

The panel held that its ultimate review of the EPA’s FIP
must await the EPA’s final action on its proposal to revise the
FIP in specific respects. Accordingly, the panel stayed the
proceedings as to evaluation of the FIP’s technical feasibility
until the administrative process was complete.

*
This summary constitutes no part of the opinion of the court. It has
been prepared by court staff for the convenience of the reader.
4 ARIZONA EX REL. DARWIN V. USEPA

COUNSEL

Peter Glaser (argued), Troutman Sanders LLP, Washington,
D.C.; Paul L. Gale, Troutman Sanders LLP, Irvine,
California; Carroll W. McGuffey III, Troutman Sanders LLP,
Atlanta, Georgia; Thomas C. Horne, Arizona Attorney
General, and James T. Skardon, Assistant Attorney General,
Phoenix, Arizona, for Petitioner State of Arizona.

Norman W. Fichthorn (argued) and Aaron M. Flynn, Hunton
& Williams LLP, Washington, D.C., for Petitioners Arizona
Public Service Company and Salt River Project Agricultural
Improvement and Power District.

Angeline Purdy (argued), United States Department of
Justice; Robert G. Dreher, Acting Assistant Attorney General,
Washington, D.C., for Respondent United States
Environmental Protection Agency.

Michael A. Hiatt (argued), Earthjustice, Denver, Colorado;
Suma Peesapati, Earthjustice, San Francisco, California, for
Respondent-Intervenors National Parks Conservation
Association and Sierra Club.

Maureen A. Scott, Matthew Laudone, and Janice M. Alward,
Phoenix, Arizona, for Amicus Curiae Arizona Corporation
Commission.

Michelle L. Wood, Phoenix, Arizona, for Amicus Curiae
Arizona Residential Utility Consumer Office.

Gordon A. Coffee, John M. Holloway III, and Stephanie B.
Sebor, Winston & Strawn LLP, Washington, D.C.; Rae
Cronmiller, Environmental Counsel, National Rural Electric
 ARIZONA EX REL. DARWIN V. USEPA 5

Cooperative Association, Arlington, Virginia, for Amicus
Curiae National Rural Electric Cooperative Association.

OPINION

BERZON, Circuit Judge:

Congress initially enacted the Clean Air Act (“the Act” or
“CAA”) in 1963 to “protect and enhance the quality of the
Nation’s air resources so as to promote the public health and
welfare and the productive capacity of its population.”
42 U.S.C. § 7401(b)(1). Later, in the Clean Air Act
Amendments of 1977, Pub. L. No. 95-95, § 128, 91 Stat. 685,
742 (current version at 42 U.S.C. § 7491), Section 169A was
added “in response to a growing awareness that visibility was
rapidly deteriorating in many places, such as wilderness areas
and national parks.” Am. Corn Growers Ass’n v. EPA,
291 F.3d 1, 3 (D.C. Cir. 2002) (internal quotation marks
omitted) (quoting Chevron U.S.A., Inc. v. EPA, 658 F.2d 271,
272 (5th Cir. 1981)).

To improve outdoor visibility, the Act as amended
“invites each State to submit to EPA a ‘State Implementation
Plan’ (‘SIP’) setting forth emission limits and other measures
necessary to make reasonable progress toward the national
visibility goal.” Nat’l Parks Conservation Ass’n v. EPA,
788 F.3d 1134, 1138 (9th Cir. 2015) (“NPCA”) (citing
42 U.S.C. §§ 7410(a), 7491(b)(2)). SIPs must include
determinations of the “best available retrofit technology”
(“BART”) to reduce emissions from certain major emission
sources, including large fossil-fuel power plants. 42 U.S.C.
§ 7491(b)(2). If a state chooses not to submit a SIP, or if
EPA disapproves a SIP in whole or in part, “the Act requires
6 ARIZONA EX REL. DARWIN V. USEPA

EPA to produce a ‘Federal Implementation Plan’ (‘FIP’) for
that State.” NPCA, 788 F.3d at 1138–39 (citing 42 U.S.C.
§ 7410(c)(1)(A)).

Arizona and the Salt River Project Agricultural
Improvement and Power District (“the State” and “SRP,”
respectively, and, collectively, “Petitioners”) petition for
review of a Final Rule (“Rule”) promulgated by EPA. The
Rule partially disapproved Arizona’s regional haze SIP
submission and promulgated a FIP in place of the
disapproved SIP elements. Approval, Disapproval and
Promulgation of Air Quality Implementation Plans, 77 Fed.
Reg. 72,512 (Dec. 5, 2012) (the “Final Rule”). In what
remains of this case,1 Petitioners challenge (1) EPA’s
disapproval of Arizona’s BART determinations, and (2) the
FIP’s replacement determinations, concerning nitrogen oxides
(NOX) emission limits at Coronado Generating Station
(“Coronado”), a two-unit, 733-megawatt coal-fueled power
plant located in Eastern Arizona.

We conclude that EPA did not act arbitrarily and
capriciously when it disapproved in part the SIP’s BART
determinations for Coronado and issued a replacement FIP as
to the disapproved portions. We therefore deny the
consolidated petitions for review.

1
See infra note 6.
 ARIZONA EX REL. DARWIN V. USEPA 7

I. STATUTORY AND REGULATORY
BACKGROUND

A. The Clean Air Act’s Visibility Protections

In enacting Section 169A2 Congress “declare[d] as a
national goal the prevention of any future, and the remedying
of any existing, impairment of visibility in mandatory class I
Federal areas which impairment results from manmade air
pollution.” § 7491(a)(1). “Class I” Federal areas include
certain national wilderness areas and national parks. Arizona
contains twelve Class I areas, the largest of which is Grand
Canyon National Park.

Section 169A seeks to reduce “regional haze,” that is,
“visibility impairment caused by geographically dispersed
sources emitting fine particles and their precursors into the
air.” Am. Corn Growers, 291 F.3d at 3 (internal quotation
marks omitted). In service of this goal, Section 169A
requires that certain sources contributing to visibility
impairment install BART—which, again, is an acronym for
“best available retrofit technology.”3 States must review all
major stationary emissions sources built between 1962 and
1977 to determine whether the source “emits any air pollutant

2
We use “Section 169A” to refer generically to the visibility
amendments, but use the codified version, 42 U.S.C. § 7491, for specific
citations.
3
We apologize for the extensive use of acronyms in this opinion and
include a brief glossary at the end to aid the reader. Environmental
litigation is awash in such alphabetical shorthand, and the “insiders”
would not know what we meant if we used other terms. We have
therefore decided against expanding the Federal Reporter 3d (F.3d, of
course, another shorthand) by stubbornly spelling out each term.
8 ARIZONA EX REL. DARWIN V. USEPA

which may reasonably be anticipated to cause or contribute
to any impairment of visibility in” any Class I area.
§ 7491(b)(2)(A). The states are then responsible for
determining the appropriate BART controls for each source.
See 42 U.S.C. § 7491(b)(2)(A), (g)(2).

EPA reviews the states’ SIP submissions, if any, for
consistency with the statute and regulations. See 42 U.S.C.
§ 7410(c)(1)(A); Oklahoma v. EPA, 723 F.3d 1201, 1204
(10th Cir. 2013). If EPA determines that a SIP does not meet
the Act’s requirements, the federal agency may itself
determine BART and impose a FIP. See 42 U.S.C.
§ 7491(b)(2)(A). More specifically, if EPA finds that a state
has not submitted a required SIP, determines that a submitted
SIP is incomplete, or disapproves a SIP in whole or in part, it
“shall promulgate a [FIP] at any time within 2 years.”
42 U.S.C. § 7410(c)(1). EPA must promulgate a FIP “unless
the State corrects the deficiency, and [EPA] approves the plan
or plan revision, before [EPA] promulgates [the FIP].” Id.

When determining BART, states or EPA must consider
five factors: “[1] the costs of compliance, [2] the energy and
nonair quality environmental impacts of compliance, [3] any
existing pollution control technology in use at the source, [4]
the remaining useful life of the source, and [5] the degree of
improvement in visibility which may reasonably be
anticipated to result from the use of such technology.”
42 U.S.C. § 7491(g)(2). Each source subject to the BART
requirement must install and operate BART “as expeditiously
as practicable but in no event later than five years after the
date of approval of a [SIP] . . . or the date of promulgation of
[a FIP].” 42 U.S.C. §§ 7491(g)(4).
 ARIZONA EX REL. DARWIN V. USEPA 9

B. EPA Regional Haze Regulations

Section 169A directed EPA to issue regulations requiring
states with Class I areas within their borders to submit SIPs
containing “emission limits, schedules of compliance and
other measures as may be necessary to make reasonable
progress toward meeting the national goal.” 42 U.S.C.
§ 7491(b)(2). EPA was also required to develop guidelines
for the states “on appropriate techniques and methods for
implementing” Section 169A. Id. § 7491(b)(1). In 1990,
Congress added Section 169B to expand the CAA’s focus to
include regional haze, see 42 U.S.C. § 7492—that is,
“visibility impairment that is caused by the emission of air
pollutants from numerous sources located over a wide
geographic area,” 40 C.F.R. § 51.301. “Section 169B
requires, among other things, that EPA undertake research to
identify ‘sources’ and ‘source regions’ of visibility
impairment in Class I areas, consider designating transport
commissions to study the interstate movement of pollutants,
and establish a transport commission for the Grand Canyon
National Park.” Am. Corn Growers, 291 F.3d at 4.

Pursuant to Sections 169A and 169B, EPA in 1999
promulgated regional haze regulations. 64 Fed. Reg. 35,714
(July 1, 1999) (codified at 40 C.F.R. § 51.300–.309). The
D.C. Circuit partially vacated those regulations in American
Corn Growers Ass’n v. EPA, 291 F.3d at 6.4 Thereafter, the
agency in 2005 promulgated new regulations, the Regional
Haze Regulations and Guidelines for Best Available Retrofit
Technology (BART) Determinations, 70 Fed. Reg. 39,104
(July 6, 2005) (“Haze Regulations”). At the same time, EPA
issued Guidelines to help states identify “BART-eligible”

4
American Corn Growers is discussed infra, at p. 21.
10 ARIZONA EX REL. DARWIN V. USEPA

sources and determine the appropriate BART for each source.
Id. at 39,156 (codified at 40 C.F.R. pt. 51, app. Y)
(“Guidelines”); see also 42 U.S.C. § 7491(b).

The Haze Regulations set a goal of achieving natural
visibility at all Class I areas by 2064. 40 C.F.R. § 51.308.
Toward that end, the Regulations direct states to submit SIPs
to EPA containing “goals (expressed in deciviews) that
provide for reasonable progress towards achieving natural
visibility conditions.” 40 C.F.R. § 51.308(d)(1). A
“deciview” is a measurement of visibility impairment. More
specifically, it “is a haze index derived from calculated light
extinction, such that uniform changes in haziness correspond
to uniform incremental changes in perception across the
entire range of conditions, from pristine to highly impaired.”
40 C.F.R. § 51.301. One deciview is the minimum visibility
impairment humans can perceive. See Guidelines at 39,120
n.32.

The SIP must also include, among other matters,
“emission limitations representing BART and schedules for
compliance with BART for each BART-eligible source that
may reasonably be anticipated to cause or contribute to any
impairment of visibility in any mandatory Class I Federal
area.” 40 C.F.R. § 51.308(e). The BART requirements apply
unless the state opts to implement an alternative emission
control measure that provides greater progress than would be
achieved through the installation of BART (commonly
referred to as “better-than BART”). See id. § 51.308(e)(2).

BART is defined as “an emission limitation based on the
degree of reduction achievable through the application of the
best system of continuous emission reduction for each
pollutant which is emitted by an existing stationary facility.”
 ARIZONA EX REL. DARWIN V. USEPA 11

Id. § 51.301. Three of the major pollutants that states must
evaluate when determining whether a source causes or
contributes to visibility impairment are sulfur dioxide
(“SO2”), nitrogen oxides (“NOX”), and particulate matter. See
Guidelines at 39,162. States must establish BART for each
pollutant. Id. at 39,163.

BART determinations for fossil-fueled power plants with
a total generating capacity greater than 750 megawatts must
comply with the Guidelines. 42 U.S.C. § 7491(b)(2). The
Guidelines set forth the following five-step process:
(1) identify all available retrofit control technologies;
(2) eliminate technically infeasible options; (3) evaluate the
control effectiveness of remaining control technologies;
(4) evaluate impacts and document the results; and
(5) evaluate the visibility impacts. Guidelines at 39,164. The
Guidelines also provide more detailed instructions for
performing each of these steps. Id. at 39,164–72. “For
sources other than 750 MW power plants . . . States retain the
discretion to adopt approaches that differ from the
[G]uidelines.” Guidelines at 39,158.

Section 309 of the Haze Regulations, 40 C.F.R. § 51.309,
allows certain western states to develop alternative visibility
improvement programs, based on the recommendations of the
Grand Canyon Visibility Transport Commission (“the
Commission”). The Commission was created to address
visibility impairment “for the region affecting the visibility of
the Grand Canyon National Park.” See 42 U.S.C. § 7492(f).
This region includes sixteen Class I areas on the Colorado
Plateau located in Arizona, Colorado, New Mexico, and Utah.
40 C.F.R. § 51.309(b). States within the relevant Transport
Region—Arizona, California, Colorado, Idaho, Nevada, New
Mexico, Oregon, Utah, and Wyoming—that submit a SIP
12 ARIZONA EX REL. DARWIN V. USEPA

compliant with all of the Commission’s recommendations are
“deemed to comply with the requirements for reasonable
progress with respect to the 16 Class I areas.” Id.
§ 51.309(a). Any covered state that elects not to submit a
Section 309 plan is “subject to the requirements of [Section
308] in the same manner and to the same extent as any State
not included within the Transport Region.” Id. Further, even
if a state submits a Section 309 SIP, it must submit a Section
308 SIP or otherwise establish “reasonable progress goals” —
including BART determinations — for any Class I areas in
the state not covered under Section 309. Id. § 51.309(g)(2).

II. PROCEDURAL HISTORY

A. The State’s SIP Submissions

The Haze Regulations set a December 17, 2007 deadline
for SIP submittals. 40 C.F.R. § 51.308(b). EPA issued a
finding on January 15, 2009, that thirty-seven states,
including Arizona, had not submitted SIPs satisfying the
Act’s visibility requirements. Finding of Failure To Submit
State Implementation Plans Required by the 1999 Regional
Haze Rule, 74 Fed. Reg. 2,392 (Jan. 15, 2009). Noting that
Arizona had “opted to develop [its SIP] based on the
recommendations of the Grand Canyon Visibility Transport
Commission” pursuant to Section 309, id. at 2,393, EPA
found that the State had failed to “submit the plan elements
required by” two provisions of the Section 309 regulations,
40 C.F.R. §§ 51.309(d)(4) and 51.309(g).5 Id. EPA further

5
Arizona’s Section 309 SIP, which Arizona submitted to EPA on
December 23, 2003, and supplemented a year later, merits further
explanation. EPA did not act upon the State’s 2003 and 2004
submissions. The Section 309 regulations were revised in October 2006,
 ARIZONA EX REL. DARWIN V. USEPA 13

explained that “[t]his finding starts the two year clock for the
promulgation by EPA of a FIP.” Id.

Although EPA determined that its January 2009 finding
triggered the Act’s two-year window, see 42 U.S.C.
§ 7410(c)(1), it did not take any further action by January
2011. Several environmental groups sued EPA in early 2011
to compel the agency to promulgate FIPs for the states
covered by the January 2009 finding. See Nat’l Parks
Conservation Ass’n v. EPA, No. 11-CV-01548 (D.D.C.).
EPA and the plaintiff groups entered into a Consent Decree
setting deadlines for EPA action for each state covered by the

following the D.C. Circuit’s decision in Center for Energy and Economic
Development v. EPA, 398 F.3d 653 (D.C. Cir. 2005). See 71 Fed. Reg.
60,612 (Oct. 13, 2006). Arizona then resubmitted its Section 309 SIP to
EPA in December 2008, acknowledging that the plan did not include
provisions required under 40 C.F.R. §§ 51.309(d)(4) and 51.309(g) as
revised. In 2013, EPA disapproved the Section 309 SIP because of these
deficiencies. 78 Fed. Reg. 48,326 (Aug. 8, 2013).

Arizona contends that, as EPA did not expressly find that its Section
309 SIP was incomplete within six months of its initial submission in
2003, the plan became complete by operation of law. Thus, it contends,
42 U.S.C. § 7410(k)(2) required EPA to disapprove its SIP within one
year, which, of course, it did not do. But, as EPA noted in its 2013
disapproval, the fact that Arizona’s submission met the completeness
criteria set out in § 7410(k) “[did] not mean that [it was] complete in the
sense that [it] contained the provisions necessary to satisfy the
requirements of [Section 309].” 78 Fed. Reg. at 48,327. The § 7410(k)
criteria “do not include the substantive provisions that a given SIP must
include to comply with the minimum requirements of the [Act].” Id.
Those provisions are instead “set out in the [Act] itself and in EPA’s
implementing regulations.” Id. EPA concluded that it could not approve
Arizona’s previously submitted SIP until it was resubmitted with valid
provisions for addressing stationary sources, provisions Arizona has
conceded were not included in the 2003 plan. Id.
14 ARIZONA EX REL. DARWIN V. USEPA

lawsuit. For Arizona, the Consent Decree required EPA by
November 15, 2012 either to approve the State’s SIP with
respect to its BART determinations or to propose a FIP.

To avoid a FIP, Arizona elected to develop a Section 308
SIP; it was submitted to EPA on February 28, 2011. The SIP
proposed progress goals and long term strategies to achieve
those goals, including BART determinations. Specifically,
the SIP included BART determinations for emission units at
three fossil fuel power plants that Arizona concluded were
BART-eligible and subject to BART: Apache Generating
Station Units 1–3, Cholla Power Plant Units 2–4, and
Coronado Generating Station Units 1–2. Only the SIP’s
determinations as to Coronado are at issue in this case.6

Overall, the State’s BART determinations for NOX, the
pollutant with which we are concerned here, consisted of
“combustion controls, either in the form of low-NOX burners
(LNB) with flue gas recirculation (FGR), or LNB with
overfire air (OFA) or separated overfire air (SOFA).” 77 Fed.
Reg. 42,834, 42,842 (July 20, 2012). For Coronado, Arizona

6
The State and other industry petitioners (Arizona Electric Power
Cooperative, Arizona Public Service Company, and PacifiCorp) initially
filed petitions for review of the Final Rule’s determinations as to Apache
and Cholla, and we consolidated our review of those petitions with this
case. On February 20, 2015, we ordered the litigation with respect to
Cholla severed and held in abeyance pending completion of state and
federal administrative proceedings that could render the petitions for
review moot. As of the publication of this opinion, those proceedings
have not yet been completed. On June 17, 2015, we dismissed the
petitions as to Apache after EPA issued a final action approving a source-
specific revision to the SIP that established a BART alternative for
Apache and withdrawing the portions of the FIP addressing BART for
Apache. See Approval and Promulgation of Air Quality Implementation
Plans, 80 Fed. Reg. 19,220 (Apr. 10, 2015).
 ARIZONA EX REL. DARWIN V. USEPA 15

determined that the proper control technology was low-NOX
burners with overfire air. Translating the chosen technology
into the resulting emission improvement, the SIP established
enforceable NOX emissions limits of 0.32 lb/mmBtu for both
units of the Coronado facility.7

B. EPA’s Actions

In July 2012, EPA proposed (1) partially to approve and
partially to disapprove the State’s BART determinations with
respect to the three power plants in its Section 308 SIP; and
(2) to promulgate a FIP for the disapproved elements.
77 Fed. Reg. at 42,834. EPA deferred taking action “on the
State’s other BART determinations or any other parts of the
SIP regarding the remaining requirements of the [Haze
Regulations].” Id. at 42,836. In a later rulemaking, EPA
approved in part and disapproved in part the remaining
portions of Arizona’s Section 308 SIP. See 78 Fed. Reg.
46,142 (July 30, 2013) (“Phase 2 Rule”).8 After a notice-and-
comment period, EPA promulgated the Final Rule challenged
here, in December 2012. 77 Fed. Reg. at 72,512.

The Final Rule approved the State’s emission limits for
SO2 and particulate matter at all the units but disapproved the
State’s emissions limits for NOX at the seven coal-fired

7
The SIP also included BART determinations for SO2 and particulate-
matter, but they are not at issue in this case.
8
Petitioners petition for review of the Phase 2 Rule in Phoenix Cement
Co. v. EPA, No. 13-73383, decided contemporaneously with this case in
a memorandum disposition.
16 ARIZONA EX REL. DARWIN V. USEPA

generating units at Apache, Cholla, and Coronado.9 Id. at
72,514. EPA explained that Arizona’s “overall approach” to
the five-step BART analysis was “generally reasonable and
consistent with the [Haze Regulations] and the BART
Guidelines.” 77 Fed. Reg. at 42,840. But it determined that
Arizona’s BART analysis suffered from several flaws,
particularly with respect to costs and visibility improvement,
that resulted in NOX control determinations “inconsistent”
with the Haze Regulations. 77 Fed. Reg. at 42,841–42; see
also 77 Fed. Reg. at 72,516–21. EPA found that the State’s
analyses with respect to SO2 and particulate matter suffered
from similar “deficiencies,” 77 Fed. Reg. at 72,517, but
nonetheless approved Arizona’s determinations for these
pollutants because the analytical flaws had no “substantive
impact on [the State’s] selection of controls.” 77 Fed. Reg.
at 42,841.

Explaining that the Consent Decree required EPA
promptly to issue a FIP for any portion of the Arizona SIP it
disapproved, the Final Rule simultaneously issued a FIP
addressing the disapproved elements. 77 Fed. Reg. at
72,567–68. EPA conducted a “new five-factor BART
analysis” of the three power plants to evaluate Arizona’s SIP
and “to document the technical basis for proposing BART
determinations in [EPA’s] FIP,” focusing in particular on
analyzing the cost controls and visibility impacts associated
with the different BART options. 77 Fed. Reg. at 42,852; see
generally 77 Fed. Reg. at 42,852–65; 77 Fed. Reg. at
72,526–61.

9
EPA approved Arizona’s BART determination for the natural gas-fired
Apache Unit 1. See 77 Fed. Reg. at 72,514.
 ARIZONA EX REL. DARWIN V. USEPA 17

Based on these analyses, EPA concluded that selective
catalytic reduction (“SCR”) with low-NOX burners and
overfire air — the most stringent available retrofit control
option — was the proper BART control for Coronado. See
77 Fed. Reg. at 42,864. EPA proposed NOX emission limits
much lower than those contained in Arizona’s SIP: 0.050
lb/mmBtu (calculated on a rolling 30-boiler-operating-day
average) for Coronado Unit 1, and 0.080 lb/mmBtu for
Coronado Unit 2.10 Id. at 42,865, tbl. 24. EPA also proposed
compliance deadlines, as well as recordkeeping and reporting
requirements, to enforce the FIP’s BART determinations.11
Id. Finally, EPA sought public comment on several aspects
of the proposed FIP. Id. at 42,835–36.

In the Final Rule, EPA revised certain elements of the
proposed FIP in response to public comments and additional
information. 77 Fed. Reg. at 72,514. Notably, EPA
weakened the final NOX emissions limits to “provide an extra
margin of compliance” and changed its methodology to
require plant-wide averaging. Id. at 72,514–15. As relevant
here, EPA changed its proposed NOX emission limits from
0.050 lb/mmBtu for Coronado Unit 1 and 0.080 lb/mmBtu for
Coronado Unit 2 to an averaged limit of 0.065 lb/mmBtu
across both units of the Coronado facility. Id. at 72,514–15,
tbl. 1. The Final Rule also extended the compliance deadlines
for installation and operation of the controls at the facilities.
See id.

10
Arizona’s SIP had adopted NOX emissions limits of 0.32 lb/mmBtu
for both units of the Coronado facility.
11
Arizona’s SIP was disapproved in part because it lacked such
deadlines and requirements. See 77 Fed. Reg. at 72,514.
18 ARIZONA EX REL. DARWIN V. USEPA

Arizona and SRP filed timely petitions for review of the
Final Rule. After the cases were consolidated, we permitted
National Parks Conservation Association and Sierra Club (the
“Respondent-Intervenors”) to intervene in the consolidated
action.

III. DISCUSSION

A. Standard of Review

Under the Administrative Procedure Act (“APA”), we
uphold a final agency action unless it is “arbitrary, capricious,
an abuse of discretion, or otherwise not in accordance with
law.” 5 U.S.C. § 706(2). To meet its regulatory obligations,
an agency must “examine the relevant data and articulate a
satisfactory explanation for its action including a rational
connection between the facts found and the choice made.”
Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Farm Mut.
Auto. Ins. Co., 463 U.S. 29, 43 (1983) (internal quotation
marks and citation omitted). An agency’s action is “arbitrary
and capricious” within the meaning of APA § 706(2) “if the
agency has relied on factors which Congress has not intended
it to consider, entirely failed to consider an important aspect
of the problem, offered an explanation for its decision that
runs counter to the evidence before the agency, or is so
implausible that it could not be ascribed to a difference in
view or the product of agency expertise.” Id. “When we
review an agency action ‘involv[ing] primarily issues of fact,’
and where ‘analysis of the relevant documents requires a high
level of technical expertise, we must defer to the informed
discretion of the responsible federal agencies.’” NPCA,
788 F.3d at 1141 (quoting Marsh v. Or. Nat. Res. Council,
490 U.S. 360, 377 (1989)).
 ARIZONA EX REL. DARWIN V. USEPA 19

The parties vigorously dispute how we should apply the
APA arbitrary and capricious standard here. EPA maintains
that in this case, as in most administrative law cases, the
standard calls for judicial deference to the federal agency as
to the decisions within its authority. Arizona (with the
support of SRP) contends, to the contrary, that Section
169A’s allocation of initial BART authority to the states
requires that we defer to “the state’s expert judgments, not to
EPA’s.” As this brief summary indicates, the parties’
disagreement concerning judicial review is grounded in
competing visions of the Act’s division of responsibility
between the states and EPA. We therefore begin our analysis
at the agency level, progressing thereafter to the question of
our review standard.

The states possess “broad authority over BART
determinations. . . . Congress intended the states to decide
which sources impair visibility and what BART controls
should apply to those sources.” Am. Corn Growers, 291 F.3d
at 8. In this connection, Section 169A of the Act gives states
the initial responsibilities of determining which sources are
subject to BART and of deciding, based on consideration of
the five statutory factors, what BART controls must be
installed and what the resulting emission limitations should
be. See 42 U.S.C. § 7491(b)(2)(A), (g)(2).

But, as the Tenth Circuit has noted, the BART
determination process “does not differ from other parts of the
[Act]—states have the ability to create SIPs, but they are
subject to EPA review.” Oklahoma, 723 F.3d at 1209; see
also N. Dakota v. EPA, 730 F.3d 750, 758 (8th Cir. 2013).
EPA may only approve SIPs, or portions of SIPs, that “meet[]
all the applicable requirements” of the Act. 42 U.S.C.
§ 7410(k)(3). Thus, while the Act “gives states discretion in
20 ARIZONA EX REL. DARWIN V. USEPA

balancing the five BART factors, it also mandates that the
state adhere to certain requirements when conducting a
BART analysis.” Oklahoma, 723 F.3d at 1208.
Consequently, when reviewing regional haze SIPs, EPA may
not approve “a BART determination that is based upon an
analysis that is neither reasoned nor moored to the [Act]’s
provisions.” N. Dakota, 730 F.3d at 761. In short, EPA is not
limited to the “ministerial” role of verifying whether a
determination was made; it must “review the substantive
content of the BART determination.” Id.

Alaska Department of Environmental Conservation v.
EPA, 540 U.S. 461 (2004) (“ADEC”), does not support the
opposite conclusion, Arizona’s contention to the contrary
notwithstanding. ADEC involved a challenge to EPA’s
authority to review states’ determinations of “best available
control technology” (“BACT”) under the Prevention of
Significant Deterioration program (“PSD”), 42 U.S.C. § 7470
et seq., a different part of the Act from the one mandating
BART. See ADEC, 540 U.S. at 468. The PSD program is
applicable to new construction of pollution emitting facilities.
See 42 U.S.C. § 7475. As with the BART process at issue
here, for the PSD program the state “exercises primary or
initial responsibility for identifying BACT in line with the
Act[].” ADEC, 540 U.S. at 484.

ADEC rejected an argument made by the petitioners in
that case similar to Arizona’s contention here—in that
instance, that EPA’s enforcement role was limited to ensuring
that the state-issued PSD permit contain a BACT
requirement. Instead, ADEC accepted EPA’s interpretation
of its authority—that EPA is authorized to review state
BACT determinations to ensure they are “reasonably moored
to the Act’s provisions.” Id. at 485, 488–89.
 ARIZONA EX REL. DARWIN V. USEPA 21

To be sure, ADEC explained, Alaska had “considerable
leeway” in its BACT determinations, and EPA was required
to accord appropriate deference to such determinations. Id.
at 490. But EPA has the authority to verify “substantive
compliance” with the Act’s BACT provisions “to guard
against unreasonable designations.” Id. at 489–90. ADEC is
thus fully consistent with our conclusion that EPA has
substantive authority to assure that a state’s proposals comply
with the Act, not simply the ministerial authority to assure
that the state has made some determination of BART.

Arizona also invokes American Corn Growers, in which
petitioners challenged aspects of EPA’s initial 1999 regional
haze rule. That case does not support Arizona’s position
regarding EPA’s SIP role either. The D.C. Circuit in
American Corn Growers disapproved EPA’s initial rule in
part because its requirements were “inconsistent with the
Act’s provisions giving the states broad authority over BART
determinations.” Am. Corn Growers, 291 F.3d at 8. But
American Corn Growers in no way suggested that once a
state has exercised its BART role by proposing a SIP, EPA
lacks authority substantively to review the SIP for
consistency with the Act.

From our determination that EPA has a substantive role
in deciding whether state SIPs are compliant with the Act and
its implementing regulations follows the conclusion that the
ordinary APA “arbitrary and capricious” judicial review
standard applies, with the requisite deference, to EPA’s
determinations. Again, neither American Corn Growers nor
ADEC supports a contrary conclusion. In particular, neither
ADEC nor American Corn Growers bolsters Arizona’s
contention that, upon judicial review, EPA bears the burden
22 ARIZONA EX REL. DARWIN V. USEPA

of proving that the State’s BART determinations are
unreasonable.

ADEC did state that, for the purposes of deciding whether
the state agency’s BACT determination in that case was
reasonable, “the production and persuasion burdens remain
with EPA.” 540 U.S. at 494. But ADEC was at that juncture
addressing a concern wholly absent here. In ADEC, the
Supreme Court held that the Act prevented EPA from
“gain[ing] a proof-related tactical advantage” by imposing a
stop-construction order rather than filing a civil action in state
court. 540 U.S. at 493–94. In that connection, ADEC noted
that “Congress nowhere suggested that the allocation of proof
burdens would differ depending upon which enforcement
route EPA selected.” Id. at 493 (emphasis added).
Petitioners in this case, however, challenge not an EPA
enforcement action but the Final Rule disapproving in part
Arizona’s SIP submission. Such prospective administrative
agency rulemaking is ordinarily reviewed under the APA’s
arbitrary and capricious standard; there is no basis for
applying a different standard here.

In sum, Section 169A gives states substantial
responsibility in determining appropriate BART controls.
EPA may not disapprove reasonable state determinations that
comply with the relevant statutory and regulatory
requirements. See Am. Corn Growers, 291 F.3d at 6–8. That
is, as ADEC put it, EPA may not “second guess” reasoned,
legally compliant state decisions. 540 U.S. at 490. But
Congress intended that EPA, not the states alone, ultimately
ensure that state determinations as to regional haze comply
with the Act, and so authorized EPA to disapprove state
“analysis that is neither reasoned nor moored to the [Act’s]
provisions.” N. Dakota, 730 F.3d at 761. Once the federal
 ARIZONA EX REL. DARWIN V. USEPA 23

agency has done so, our role as the reviewing court remains
what it always is when substantively reviewing agency action
under the APA—deciding, with appropriate deference to the
federal agency, whether the agency’s action was “arbitrary,
capricious, an abuse of discretion, or otherwise not in
accordance with law.” 5 U.S.C. § 706(2).

B. Partial Disapproval of the SIP’s BART
Determinations

We now turn to Petitioners’ challenges to EPA’s partial
disapproval of Arizona’s SIP as it applies to the Coronado
BART determinations.

1. Arizona’s procedural claim

Before addressing Petitioners’ substantive arguments, we
address Arizona’s procedural argument. Arizona maintains
that, in acting only on the BART determinations when it
promulgated the Final Rule, deferring action on the rest of the
SIP, EPA “misconstrued its statutory authority and acted in
an arbitrary fashion.” Specifically, Arizona contends that
EPA cannot properly evaluate BART determinations
separately from the broader reasonable progress analysis,
because BART determinations are just one aspect of
achieving Section 169A’s overall “reasonable progress” to
the natural visibility goal. 42 U.S.C. § 7491(b)(2).

That is not so. There is no requirement that EPA approve
or disapprove a SIP submittal in a single action. To the
contrary, the Act expressly permits EPA to approve or
disapprove a SIP “in part.” 42 U.S.C. § 7410(c)(1), (k)(3);
see also Ass’n of Irritated Residents v. EPA, 423 F.3d 989,
997 (9th Cir. 2005).
24 ARIZONA EX REL. DARWIN V. USEPA

The EPA rule partially disapproving a regional haze SIP
which was recently upheld by Oklahoma, 723 F.3d at 1206,
for example, “t[ook] no action on whether Oklahoma has
satisfied the reasonable progress requirements.” 76 Fed. Reg.
81,728, 81,730 (Dec. 28, 2011). Under the Haze Regulations,
SIPs must contain reasonable progress goals, 40 C.F.R.
§ 51.308(d)(1), as well as source-specific BART
determinations, id. § 51.308(e)(1).12 Of course, the
implementation of the BART determinations will ultimately
contribute toward meeting the reasonable progress goals. But
the Act sets out standards for BART that are freestanding,
source-by-source, and not dependent on the long term
visibility goals identified. EPA did not act arbitrarily by
considering Arizona’s BART determinations separately from
the State’s reasonable progress analysis.

2. Arizona’s substantive claims

Turning to the substance of the Final Rule: EPA found
that the SIP’s “overall approach” was “generally reasonable
and consistent” with the Haze Regulations and the
Guidelines.13 77 Fed. Reg. at 42,840. But it concluded that
the State’s BART determinations were deficient in three
particular respects. First, Arizona’s control cost calculations

12
As Arizona argues, in “some circumstances” BART controls may not
be “necessary to make reasonable progress.” But that is so when states
elect to implement an “emissions trading program or other alternative
measure,” that result in “greater reasonable progress” than BART. Id.
§ 51.308(e)(2) (emphasis added). Here, the State did not submit an
alternative measure under § 51.308(e)(2), choosing instead to conduct a
source-specific BART analysis.
13
Petitioners do not challenge the Haze Regulations or the Guidelines,
only their implementation in the Final Rule here at issue.
 ARIZONA EX REL. DARWIN V. USEPA 25

were not performed in accordance with the Guidelines and
were otherwise unreasonable. See id. at 42,841; 77 Fed. Reg.
at 72,516–18, 72,566. Second, Arizona did not properly
evaluate the visibility improvements to all Class I areas. See
77 Fed. Reg. at 42,841–42; 77 Fed. Reg. at 72,519, 72,566.
And third, Arizona inadequately explained its consideration
of the BART factors. See 77 Fed. Reg. at 42,841, 42,846; 77
Fed. Reg. at 72,517, 72,566. Because of these deficiencies,
EPA partially disapproved Arizona’s BART determinations,
including those pertaining to Coronado’s NOX emission
controls.

Arizona and SRP contend that each of EPA’s conclusions
is legally flawed and unsupported by the record.
Consequently, they contend, the agency’s partial disapproval
of the SIP was arbitrary and capricious. We disagree. EPA’s
conclusions concerning the State’s BART analysis and
determinations were well explicated, carefully grounded in
the administrative record, and analytically reasonable, and so
properly support its disapproval of Arizona’s NOX BART
determinations for Coronado.

(i) The SIP’s Cost Analysis

As one of the BART factors, states must consider the
“costs of compliance.” 42 U.S.C. § 7491(g)(2); 40 C.F.R.
§ 51.308(e)(1)(ii)(A). “States have flexibility in how they
calculate costs.” Haze Regulations, 70 Fed. Reg. at 39,127.
A state’s cost calculations are critical to determining a BART
control’s “cost effectiveness,” where “‘effectiveness’ is
measured in terms of tons of pollutant emissions removed,
and ‘cost’ is measured in terms of annualized control costs.”
Id. at 39,167.
26 ARIZONA EX REL. DARWIN V. USEPA

After identifying control alternatives and achievable
emissions performance levels, states are directed to “develop
estimates of capital and annual costs.” Id. at 39,166. “The
basis for equipment cost estimates also should be
documented, either with data supplied by an equipment
vendor . . . or by a referenced source (such as the [Office
of Air Quality Planning and Standards] Control Cost
Manual . . . ).” Id.

The Guidelines instruct that the Cost Manual “addresses
most control technologies in sufficient detail for a BART
analysis,” and that “cost estimates should be based on the
[Cost Manual] where possible” to “maintain and improve
consistency.” Id. States are allowed by the Guidelines to
include “additional information” — such as “any information
supplied by vendors that affects your assumptions regarding
purchased equipment costs, equipment life, [or] replacement
of major components” — in their cost calculations, but
require them to provide documentation for any “element of
the calculation that differs from the . . . Cost Manual.” Id.
n.15. Finally, the Guidelines indicate that “[t]he cost analysis
should also take into account any site-specific design or other
conditions . . . that affect the cost of a particular BART”
option. Guidelines at 39,166.

In its proposed rule, EPA found “certain aspects” of
Arizona’s cost calculations “inconsistent” with the Guidelines
and Cost Manual and “disagree[d] with the manner in which
[Arizona] interpreted the cost-related information included in
its [] SIP.” 77 Fed. Reg. at 42,841. With regard to Coronado,
specifically, EPA noted in its proposed rule that SRP
“provided summaries of total control costs, such as total
annual operating and maintenance costs and total annualized
capital cost, but did not provide cost information at a level of
 ARIZONA EX REL. DARWIN V. USEPA 27

detail that included line item costs.” Id. at 42,850. This
omission meant that SRP did not provide Arizona “with
control cost calculations at a level of detail that allowed for
a comprehensive review.” Id. at 42,851. As a result, EPA
explained, it “[did] not believe that [Arizona] was able to
evaluate whether SRP’s control costs were reasonable.” Id.
Arizona’s BART analysis was therefore inadequate because
it “did not properly consider the costs of compliance for each
control option.” Id.

We conclude that EPA’s disapproval of the cost analysis
underlying Arizona’s BART determination for Coronado on
that basis was not “arbitrary, capricious, [or] an abuse of
discretion.” 5 U.S.C. § 706(2)(A). Arizona simply relied on
the cost data provided by SRP, despite the fact that the data
failed to include sufficient detail for the State meaningfully
to analyze the reasonableness of the costs of various control
alternatives. States are required by statute to consider “costs
of compliance” in making BART determinations. 42 U.S.C.
§ 7491(g)(2). When they are not presented with enough data
to do so, EPA may reasonably conclude that their analysis is
28 ARIZONA EX REL. DARWIN V. USEPA

inadequate.14 EPA’s decision to do so was not arbitrary or
capricious.

(ii) The SIP’s Visibility Analysis

As part of its BART analysis, a state must analyze “the
degree of improvement in visibility which may reasonably be
anticipated to result from the use” of alternative control
technologies. 40 C.F.R. § 51.308(e)(1)(ii)(A). EPA found no
problems with the “technical adequacy of [Arizona’s
visibility] modeling.” 77 Fed. Reg. at 72,519. Rather, EPA
found Arizona’s interpretation of the visibility modeling for
all three plants “problematic.” Id. The problems, EPA
contends, resulted in Arizona understating the visibility
benefits associated with installing SCR at Coronado. We
conclude that EPA’s assessment of Arizona’s visibility
analysis considered the appropriate factors rationally, and so
defer to its conclusions. See Motor Vehicle Mfrs., 463 U.S.
at 43.

14
EPA also disapproved the cost analysis for failing to use the
“overnight method” required by the Cost Manual. The “overnight”
method “treats the costs of a project as if the project were completed
‘overnight,’ with no construction period and no interest accrual.” 77 Fed.
Reg. at 72,530. “Since assets under construction do not provide service
to current customers,” utilities for ratemaking purposes use an alternative,
“levelized” methodology, to “capitalize[] the interest and return on equity
that would accrue over the construction period and adds them to the rate
base when construction is completed and the assets are used.” Id.

Because EPA had a sufficient alternative basis for disapproving the
Coronado cost analysis, we do not here decide whether it could require
Arizona to employ the overnight method. We discuss EPA’s use of the
overnight method in its FIP infra, at 39–40.
 ARIZONA EX REL. DARWIN V. USEPA 29

For Coronado, Arizona used a “visibility index”
averaging the visibility benefits at the closest nine Class I
areas, but did not evaluate such benefits separately at the most
impacted Class I area, the Gila Wilderness Area. 77 Fed.
Reg. at 72,519; see also 77 Fed. Reg. at 42,850–51. EPA’s
regulations “do not prescribe a particular approach to
calculating or considering visibility benefits across multiple
Class I areas,” 77 Fed. Reg. at 42,841; states have the
“flexibility to assess visibility improvements due to BART
controls by one or more methods,” Guidelines at 39,170. The
indexing approach therefore “could be acceptable in itself as
part of assessing multiple area impacts and improvements.”
77 Fed. Reg. 72,519. But, EPA concluded, “without any
consideration of particular area improvements, the averaging
process causes especially large benefits at some individual
areas to be diluted or lost, effectively discounting some of the
more important effects of the controls.” Id. (emphasis
added).

Moreover, regardless of the methodology used, EPA
maintains, Arizona’s visibility analysis in its SIP was
unreasonable because it used “two contrasting, yet
equally incomplete, approaches to assessing visibility
improvements.” Arizona used a visibility index average to
analyze visibility benefits at Coronado, but its analyses for
Apache and Cholla considered visibility improvements “only
at the single Class I area with the greatest modeled impact
from a facility,” rather than at all impacted Class I areas.
77 Fed. Reg. at 72,519. That is, the cumulative averaging
approach taken by Arizona in its analysis for Coronado “is
counter to [Arizona’s] emphasis elsewhere in the SIP on the
importance of considering the visibility improvement at the
single area having the largest impact from a given facility.”
Id. The upshot is the appearance that the State selectively
30 ARIZONA EX REL. DARWIN V. USEPA

chose for each plant a methodology that minimized the
visibility improvement achieved by the more stringent
emission controls at each location.

Arizona made no attempt in its SIP, nor in its briefing in
this appeal, to counter this appearance by explaining why it
chose differing approaches to visibility analysis for different
facilities. SRP may be correct that “[t]he Guidelines allow
states to use either or both approaches.” But, as described
above, a state must include in its SIP “an explanation of the
CAA factors that led [the State] to choose that option over
other control levels.” Guidelines at 39,170–71 (emphasis
added). Adopting inconsistent — indeed, contradictory —
approaches without providing any explanation for that
decision frustrated EPA’s ability to “review the substantive
content of the BART determination.” N. Dakota, 730 F.3d at
761.

Arizona also contends that the outcome of its BART
determinations would not have changed even if it had adopted
the approach to visibility analysis EPA prescribed. The
visibility improvements resulting from installing SCR, the
State maintains, would in any event be “imperceptible” to the
human eye.

EPA expressly, and reasonably, rejected this argument
when it promulgated the Haze Regulations and Guidelines in
2005:

Even though the visibility improvement from
an individual source may not be perceptible,
it should still be considered in setting BART
because the contribution to haze may be
significant relative to other source
 ARIZONA EX REL. DARWIN V. USEPA 31

contributions in the Class I area. Thus, we
disagree that the degree of improvement
should be contingent upon perceptibility.
Failing to consider less-than-perceptible
contributions to visibility impairment would
ignore the CAA’s intent to have BART
requirements apply to sources that contribute
to, as well as cause, such impairment.

70 Fed. Reg. at 39,129.

In sum, EPA rationally determined that Arizona’s BART
visibility analysis for Coronado was unsupported by
explanation and inconsistent with the CAA and its
regulations. We defer to its conclusions.

(iii) The SIP’s Choice of BART

EPA’s Guidelines require states to support their BART
determinations with “documentation for all required
analyses,” including explanations of their BART five-factor
analysis. 40 C.F.R. § 51.308(e)(1). Further, the Guidelines
indicate that states “should provide a justification for
adopting the technology . . . select[ed] as the ‘best’ level of
control, including an explanation of the [Act’s] factors that
led [the State] to choose that option over other control
levels.” Guidelines at 39,170–71 (emphasis added). A state
need not perform this analysis if a source already has, or has
committed to installing, the most stringent controls available.
Id. at 39,165. Otherwise, as EPA explained in proposing the
Rule here at issue, “[s]tates are free to determine the weight
and significance assigned to each factor, but must consider all
five factors and provide a reasoned explanation for adopting
[BART].” 77 Fed. Reg. at 42,838 (emphasis added).
32 ARIZONA EX REL. DARWIN V. USEPA

In the Final Rule, EPA concluded that, although Arizona
“presented information relevant to each of the BART factors”
and “expressly stated” that it had considered those factors, it
did not “provide[] an explanation regarding how this
information was used to develop its BART determinations.”
77 Fed. Reg. at 72,517; see also id. at 72,566 (“[Arizona] has
not demonstrated that it actually took into consideration the
BART factors in making its determinations[.] . . . [I]t gave no
explanation or rationale for how it reached a determination
based on that information.”). More specifically, EPA found
that Arizona did not discuss how the results of the visibility
index were weighed against the other BART factors for
Coronado. Id. at 72,518; 77 Fed. Reg. at 42,851. Further,
EPA noted that while the SIP includes cost data, it “provides
no explanation regarding how, or even if, th[e] cost
information was used in arriving at its NOX BART
determinations.” 77 Fed. Reg. at 72,517. Indeed, “[i]n the
case of . . . Coronado, the . . . SIP does not analyze th[e] cost
information in even a qualitative manner.” Id.

A review of Arizona’s BART Technical Support
Document supports EPA’s analyses with regard to Coronado.
Arizona’s ultimate determination was that, “[a]fter reviewing
the BART analysis provided by the company, and based upon
the information above . . . BART control at [Coronado] for
NOX is . . . Low NOX burners with OFA” with an emission
rate of 0.32 lbs/mmBtu. Before announcing that decision,
Arizona provided several charts of data concerning the
various controls’ cost-effectiveness and visibility impacts.
But, having done so, it provided no reasoning or rationale to
justify its ultimate BART selection. There was simply
no attempt made to explain why the State chose one
control technology over another, or how it evaluated the
various BART factors (i.e., cost-effectiveness, visibility
 ARIZONA EX REL. DARWIN V. USEPA 33

improvement, energy and non-air quality environmental
impacts, existing pollution control technology in use, and the
remaining useful life of the source), either individually or in
combination.

Indeed, Arizona does not meaningfully contest EPA’s
adverse characterization of its BART analysis. Rather, it
contends that “no great explanation is required to understand
the State’s decision not to spend hundreds of millions of
dollars of its citizens’ money for an imperceptible
improvement in visibility.” Yet, under the Act and its
implementing regulations, states are required in SIPs to
explain the choice of BART, taking into account not only cost
and visibility improvement, but also the three other BART
factors. See 40 C.F.R. § 51.308(e)(1)(ii)(A); Guidelines at
39,170–71.

We recently invalidated a FIP in part because EPA’s
conclusory cost-benefit analysis “fail[ed] to reveal to a reader
how EPA determined that the cost of controls were not
justified.” NPCA, 788 F.3d at 1145. The same failure here,
albeit by a state in its SIP, fares no better. Although the Act
affords states significant discretion in determining the
appropriate levels of BART controls, EPA must review
whether a state’s determinations comply with the statute and
its rules. See N. Dakota, 730 F.3d at 761; Oklahoma,
723 F.3d at 1209. Just as we could not in NPCA review
EPA’s cost/benefit analysis absent any coherent agency
analysis, EPA reasonably determined that it could not
meaningfully review Arizona’s parallel determination,
because the State did not provide an adequate explanation of
its underlying analysis, if any. Cf. 40 C.F.R.
§ 51.308(e)(1)(ii)(A); Guidelines at 39,170–71.
34 ARIZONA EX REL. DARWIN V. USEPA

In sum, EPA’s conclusion that Arizona did not adequately
explain its NOX BART determinations provides reasonable
support for its partial disapproval of Arizona’s SIP. We
therefore defer to EPA’s determination.

(iv) Conclusion

EPA reasonably concluded that Arizona’s cost and
visibility impact analyses for Coronado suffered from
significant analytical defects and that the SIP did not provide
a reasoned explanation of the bases for the ultimate BART
determination for Coronado. Although Section 169A affords
the states substantial authority to determine BART controls,
the combination of these defects provided EPA reasonable
grounds upon which to disapprove the Arizona’s BART
determinations as to NOX emissions limits at Coronado. Its
partial disapproval of the SIP in this respect was not arbitrary
or capricious.

C. EPA’s Promulgation of the FIP

To remedy the deficiencies it identified in its partial
disapproval of Arizona’s SIP, EPA in the same Final Rule
promulgated a FIP. As relevant here, EPA concluded that
selective catalytic reduction with low-NOX burners and
overfire air, “the most stringent available control option” for
NOX emissions, was BART for Coronado. 77 Fed. Reg. at
42,864. It thus proposed NOX emission limits of 0.05
lb/mmBtu, based on a rolling 30-boiler-operating day
average,15 for Coronado. Id. at 42,865. After receiving

15
“Boiler operating day means a 24–hour period between 12 midnight
and the following midnight during which any fuel is combusted at any
time in the steam-generating unit. It is not necessary for fuel to be
 ARIZONA EX REL. DARWIN V. USEPA 35

comments on its proposed rule, EPA’s Final Rule imposed a
FIP containing revised NOX emission limits of 0.065
lb/mmBtu, measured as an average of the two Coronado
generating units. See 77 Fed. Reg. at 72,514–15. This final
emission limit is, of course, very significantly more stringent
than the 0.32 lb/mmBtu NOX limit proposed by Arizona in its
SIP. As required by the Guidelines, EPA explained its
general analytical approach and BART determinations at
substantial length in the Final Rule. See id. at 72,526–35.

SRP presses a number of substantive challenges to the
FIP’s emission limit, as well as to EPA’s analysis underlying
that determination. Arizona, on the other hand, challenges
EPA’s decision to promulgate a FIP in the same rule in which
it disapproved in part Arizona’s SIP. We address these
challenges in turn.

1. The FIP’s BART Determinations

Under Section 169A and the Haze Regulations, EPA must
perform the same BART analysis when promulgating a
regional haze FIP as that performed by states in developing
SIPs. See 42 U.S.C. § 7491(b)(2)(A), (g)(2); see also 40
C.F.R. § 51.308(e)(1)(ii)(A)–(B). Challenges to the FIP are
reviewed “under the same arbitrary and capricious standard
. . . used to evaluate the EPA’s rejection of the SIP.”
Oklahoma, 723 F.3d at 1215. But this review requires a
“slightly different perspective,” as we “evaluat[e] the EPA’s
own choices under the guidelines, as opposed to evaluating its

combusted the entire 24–hour period.” 40 C.F.R. § 60.41. At the end of
each boiler operating day, emissions are measured and a new 30-day
rolling average is calculated. See 40 C.F.R. § 60.48Da(b).
36 ARIZONA EX REL. DARWIN V. USEPA

choice to reject [Arizona’s] SIP under the guidelines.” Id.
(emphasis added).

SRP contends that EPA’s cost and visibility analyses
contained in its FIP were arbitrary and capricious. It also
maintains that, for a number of reasons, the FIP’s NOX
emission limits are neither achievable nor reasonable.
Consequently, it argues, the FIP’s BART determinations for
Coronado were, on the whole, arbitrary and capricious. For
the reasons set forth below, we in the main do not agree, but
we leave one issue open because EPA is reconsidering it.

a. EPA’s Visibility Analysis

SRP first disputes EPA’s “cumulative” approach. EPA
estimated the visibility improvements that would occur at
each of the Class I areas potentially impacted by Coronado’s
emissions and then aggregated those improvements. SRP
contends this analysis resulted in “a large deciview number”
that does not represent the actual perception of visibility
conditions at any particular Class I area. SRP’s challenge to
EPA’s visibility improvement analysis suffers from two
substantial defects.

First, EPA considered the “cumulative visibility
improvement” resulting from various control technologies
“[a]s a supplement” to considering deciview improvements
at individual Class I areas. For Coronado, specifically, EPA
explained that modeling showed that SCR control technology
would result in visibility benefits at each of eleven Class I
areas—including the Gila Wilderness Area, which EPA
faulted Arizona for failing to consider—as well as on a
cumulative basis. See 77 Fed. Reg. at 42,863 tbl.23. In
response to Petitioners’ comments critiquing EPA’s
 ARIZONA EX REL. DARWIN V. USEPA 37

cumulative approach, EPA explained in the Final Rule that
“[t]he approach is simply one way of assessing improvements
at multiple areas, for consideration along with other visibility
metrics.” 77 Fed. Reg. at 72,532. SRP’s claim that EPA
“focused on this [cumulative] methodology almost
exclusively” is simply not supported by the record.

Second, SRP’s insistence on “human perception” as the
determinative “cornerstone” for the BART determinations for
each individual source is overstated. As discussed above,
when promulgating the BART Guidelines, EPA explicitly
disagreed “that the degree of improvement should be
contingent upon perceptibility” when determining BART for
an individual source. 70 Fed. Reg. at 39,129.

EPA’s Final Rule provided a fully adequate explanation
of its application of the deciview concept in the FIP. Again,
one deciview is the minimum visibility difference people can
perceive. See Guidelines at 39,120 n.32. The Guidelines
suggest that states use a minimum threshold of 0.5 deciviews
to determine whether a source is subject to BART controls.
EPA explained in its Final Rule, however, that “[s]maller
improvements from controls should be considered in BART
determinations, since they can be beneficial in considering
effects from controls on multiple sources.” 77 Fed. Reg. at
72,533 (emphasis added).

In conclusion, EPA’s visibility improvement assessment
was consistent with the statute and regulatory requirements,
and supported by the record.
38 ARIZONA EX REL. DARWIN V. USEPA

b. EPA’s Cost Analysis

SRP also challenges EPA’s cost analysis, arguing that it
“diverged” from the Guidelines and was “inadequate” to
support its BART determinations for NOX emissions.
Additionally, SRP argues that EPA’s cost assessment was
flawed “because [EPA] insisted on rigid adherence to the
[Cost Manual] irrespective of site-specific costs.”

EPA explains that it used the “air pollution control cost
development” component of the Integrated Planning Model
(“IPM”) to develop its cost estimates. See 77 Fed. Reg. at
72,530. IPM, a model of the United States electric power
sector, “relies upon a very large number of data inputs and
provides forecasts” of costs and other variables relevant to
decisionmaking in that sector. Id. SRP maintains that by
using IPM cost estimates, EPA ignored the Coronado
facility’s “site-specific characteristics,” and so understated
the actual costs of compliance for installing controls at
Coronado.

In the Final Rule, however, EPA explained that it did not
rely on IPM as a whole, but rather on “one component of
IPM, specifically, the component that develops the costs of
air pollution control technologies.” Id. As EPA explained,
that cost development methodology was not “generic” or
“generalized.” Id. Rather, it was grounded in databases of
actual SCR projects from 2004, 2006, and 2009. Id. At the
same time, EPA recognized that “a costly engineering
evaluation that included site visits would potentially produce
a more refined cost estimate that could be considered more
site-specific than our own.” Id.
 ARIZONA EX REL. DARWIN V. USEPA 39

In addition, in response to Petitioners’ public comments
contending that EPA failed to consider site-specific
information, EPA conducted supplemental cost analyses that
relied upon “cost estimates provided by SRP.” See id. at
72,558–60. Based on this supplemental cost analysis, EPA
concluded that “the cost-effectiveness values of SCR . . . [are]
not . . . cost-prohibitive” for Coronado. Id. at 72,560.

Not satisfied, SRP contends that in its supplemental
analysis, “EPA excluded costs it deemed inconsistent with the
[Cost Manual]” such as Allowance for Funds Utilized During
Construction (“AFUDC”).16 This argument restates
Petitioners’ objections to EPA’s reliance on the overnight
costing methodology when it partially disapproved Arizona’s
SIP. See supra note 14. EPA’s use of such a methodology in
its own FIP’s cost analysis is, without doubt, reasonable. See
77 Fed. Reg. at 72,518.

In rejecting Arizona’s SIP, EPA explained that the use of
the overnight method was “crucial to [its] ability to assess the
reasonableness of the costs of compliance.” Id. The agency
went on:

A proper evaluation of cost-effectiveness
allows for a reasoned comparison not only of
different control options for a given facility,
but also of the relative costs of controls for
similar facilities. If the cost-effectiveness of
a control technology for a particular facility is
outside the range for other similar facilities,

16
“AFUDC primarily represents a tool for utilities to capture their cost
of borrowing and return on equity during construction periods.” 77 Fed.
Reg. at 72,531.
40 ARIZONA EX REL. DARWIN V. USEPA

the control technology may be rejected as
not cost-effective. . . . Without an
‘apples-to-apples’ comparison of costs, it is
impossible to draw rational conclusions about
the reasonableness of the costs of compliance
for particular control options. Use of the
[Cost Manual] methodology is intended to
allow a fair comparison of pollution control
costs between similar applications for
regulatory purposes.

Id. EPA concluded that “it is reasonable for us to insist that
the [Cost Manual] methodology be observed in the cost
estimate process.” Id. Accordingly, it rejected comments
that items like AFUDC should have been incorporated into its
cost analysis, as they were “inconsistent with [the Cost
Manual] methodology.” Id. at 72,531.

EPA’s analysis is reasonable. The purpose of the cost
analysis required as part of a BART determination is to foster
comparison of the cost of the visibility improvements enabled
by various control technologies. As EPA’s comments
indicate, cross-facility comparisons of similar sources with
regard to the cost-effectiveness of a given control option aid
in determining cost-effectiveness at a specific source.
Control options are likely to impact similar sources similarly;
comparisons assure that the cost and benefit figures used for
a particular site are realistic, rather than inflated in one
direction or another. Consideration of AFUDC would not
further this inquiry, as AFUDC is ultimately reflective of the
implementing entity’s financial and logistical situations,
grounded in past decisions and in the company’s financial
policies and attitudes, not of the hard costs of the equipment
and construction, which should be consistent across sites.
 ARIZONA EX REL. DARWIN V. USEPA 41

While AFUDC and similar concepts are relevant for sales and
ratemaking, including them would undermine the sort of
“apples-to-apples” comparison that EPA asserts is necessary
as part—but only part—of assessing the control options.

This approach is consonant with the Guidelines, which
specifically advise that “reasonable range[s]” for cost
effectiveness are those that are “consistent with the range of
cost effectiveness values used in other similar permit
decisions over a period of time.” Guidelines at 39,168; see
also Oklahoma, 723 F.3d at 1213 (“The guidelines say that
states should follow the manual’s methodology so that
projects can be more easily compared.”). Moreover, adopting
a costing methodology which focuses on achieving
consistency and facilitating comparisons aligns with the CAA
itself, which empowers EPA to promulgate national
regulations concerning BART determinations. See 42 U.S.C.
§ 7491(a)(4), (b)(1).

Accordingly, we reject SRP’s argument that the FIP’s
underlying cost analysis was arbitrary and capricious.

c. Achievability of the FIP’s NOX emission limits for
Coronado

The Haze Regulations provide that the BART
determination “must be based on an analysis of the best
system of continuous emission control technology available
and associated emission reductions achievable for each
BART-eligible source that is subject to BART.” 40 C.F.R.
§ 51.308(e)(1)(ii)(A) (emphasis added). The reviewing
authority should “take into account the most stringent
emission control level that the technology is capable of
achieving,” by considering “recent regulatory decisions and
42 ARIZONA EX REL. DARWIN V. USEPA

performance data (e.g., manufacturer’s data, engineering
estimates and the experience of other sources).” Guidelines
at 39,166.

SRP argues that the FIP’s NOX emission limit for
Coronado — 0.065 lb/mmBtu averaged across the facility —
is not achievable. More specifically, it argues that (1) the
emission limits are technically infeasible; and (2) EPA did
not take into account the Consent Decree binding Coronado
Unit 2 when formulating the emission limits.

Both of these arguments will be rendered moot if EPA’s
recent action proposing to revise the FIP’s NOX emission
limit for Coronado results in a final revised FIP consistent
with the proposal. See 80 Fed. Reg. 17,010 (March 31,
2015).17 The proposed revision would replace the facility-
wide compliance method with “a unit-specific compliance
method for determining compliance with . . . BART []
emission limits for nitrogen oxides,” and would establish
“unit-specific limit[s]” of 0.065 lb/mmBtu for Coronado Unit
1 and 0.080 lb/mmBtu for Coronado Unit 2. Id.

SRP’s technical feasibility argument was largely based on
the assertion that a 0.050 lb/mmBtu emissions limit — which
SRP contended Coronado Unit 1 would have to satisfy —
was “infeasible” for SCR retrofits to coal-fired electric
generating units. Even though EPA based its BART
determinations for Coronado on modeling showing that SCR
controls could achieve a 0.050 lb/mmBtu NOX emission rate
with an 80%–90% control efficiency, see 77 Fed. Reg. at
42,853, and has imposed a NOX emission limit in the 0.050

17
EPA has recently informed us that it expects to issue the final revised
FIP by March 2016.
 ARIZONA EX REL. DARWIN V. USEPA 43

lb/mmBtu range in other, similar rulemakings, see 77 Fed.
Reg. at 72,544 tbl.7, the proposed revised FIP’s limits for
Coronado Unit 1 are now 0.065 lb/mmBtu.18 Likewise,
although SRP initially argued that the FIP’s emission limits
were inconsistent with the limits prescribed by a pre-existing
Consent Decree as to Coronado Unit 2, EPA’s proposed FIP
revision establishes the same emissions limit as that
prescribed by the Consent Decree: 0.080 lb/mmBtu. See
80 Fed. Reg. at 17,016–18. Consequently, these arguments,
whatever merit they may have had as to the original FIP, will
not carry force if the proposed revised FIP is adopted.

Additionally, SRP argues that the FIP’s emissions limits
are in conflict with EPA’s own “presumptive” limits. The
BART Guidelines provide a list of “presumptive NOX limits,”
based on boiler and coal type, for certain coal-fired
generating units operating without post-combustion controls.
Guidelines at 39,171. Under the Guidelines, the reviewing
authority “should require such [facilities] to meet the
[presumptive] NOX emission limits, unless [it] determine[s]
that an alternative control level is justified based on
consideration of the statutory factors.” Id. The presumptive
NOX limits for Coronado are 0.32 lb/mmBtu for bituminous
coal and 0.23 lb/mmBtu for sub-bituminous coal. Id. tbl.1.
Arizona’s SIP proposed a NOX emission limit for Coronado
within the presumptive limit range, while the FIP imposes a
far lower facility-wide limit. SRP argues that EPA, unlike
Arizona, did not consider the presumptive limits, and
therefore violated the Guidelines.

18
A 0.065 lb/mmBtu emissions limit may also be, in SRP’s view,
technically infeasible, and SRP is entitled to petition for review of EPA’s
action finalizing the FIP revision when it is issued. See 42 U.S.C.
§ 7607(b).
44 ARIZONA EX REL. DARWIN V. USEPA

As the Final Rule explains, EPA did consider the
presumptive limits but found there is “no single presumptive
NOX limit that applies to any of these units,” as each of the
facilities “historically burned both bituminous and sub-
bituminous coal.” 77 Fed. Reg. at 72,529. Accordingly, EPA
instead “considered the technological basis for presumptive
NOX BART limits . . . as part of the five-factor analysis [it]
performed for each facility.” Id.

In any event, SRP’s argument that the “law requires [the
presumptive] limits to be taken into account in any BART
determination” is belied by the Guidelines. The presumptive
emission limits are “rebuttable” and “do[] not preclude states
or EPA from setting limits that differ from those
presumptions.” 77 Fed. Reg. at 72,529. Instead, the
Guidelines expressly allow for an alternative control level to
be formulated based on the statutory factors, provided that the
alternative limits are based on a reasoned BART analysis.
70 Fed. Reg. at 39,171. Moreover, the presumptive emission
limits are presumed to be cost-effective, not presumed to be
BART in every case. See 77 Fed. Reg. at 51,620, 51,633
(Aug. 24, 2012); 77 Fed. Reg. 14,604, 14,665 (Mar. 12,
2012).

In sum, EPA acted reasonably in departing from
Guidelines’ rebuttable presumptive limits. But, because EPA
has not yet completed its proposed revised FIP, we decline to
rule on the reasonableness of its emissions limits, as they are
likely to be altered. This aspect of these proceedings is
therefore stayed until EPA concludes the administrative
process and issues its final revised FIP.
 ARIZONA EX REL. DARWIN V. USEPA 45

2. Simultaneous Disapproval of a SIP and
Promulgation of a FIP

Apart from its substantive challenges to the FIP, Arizona
contends that EPA procedurally erred in promulgating the FIP
in the same rule as its partial disapproval of the SIP. That is
not so.

Under the Act, EPA shall promulgate a FIP “at any time
within 2 years” after EPA disapproves a SIP in whole or in
part or finds that a state has not made a required submission.
42 U.S.C. § 7410(c) (emphasis added). A state may forestall
the promulgation of a FIP if it “corrects the deficiency, and
[EPA] approves the plan or plan revision, before [EPA]
promulgates such [FIP].” Id. Furthermore, a state may
submit a SIP revision to EPA at any time, and EPA must act
on it within twelve months of submission. Id. § 7410(k).

The Final Rule’s partial disapproval of Arizona’s BART
determinations constituted a trigger under the Act for
promulgating a FIP replacing those elements. The Act
expressly authorizes EPA to promulgate a FIP “at any time”
within two years of disapproving a SIP. “At any time,” of
course, includes simultaneously with the SIP’s disapproval.
See Oklahoma, 723 F.3d at 1223.

Arizona ultimately recognizes that EPA had the authority
to promulgate the FIP simultaneously with its partial
disapproval, but maintains that EPA did not realize that it had
discretion to provide Arizona up to two years to correct any
deficiencies. EPA, Arizona posits, harbored the erroneous
belief that its obligations under the Consent Decree required
it to act when it did.
46 ARIZONA EX REL. DARWIN V. USEPA

Some procedural background is required to understand
the thrust of Arizona’s argument. After EPA did not
promulgate FIPs within two years of its January 2009 finding
that many states, including Arizona, had not submitted
required SIPs, it entered into a Consent Decree requiring the
federal agency to approve a SIP or promulgate a FIP by
November 15, 2012. See supra, p. 13–14. In the Final Rule,
EPA stated that the Consent Decree “required [it] to issue a
FIP for any portion of the Arizona SIP that we cannot
approve.” 77 Fed. Reg. at 72,569. EPA further explained
that “while . . . in the absence of an expired statutory duty and
a court-ordered deadline to issue a FIP, it would be preferable
for us to give Arizona additional time to revise its Regional
Haze SIP prior to promulgation of a FIP, we simply do not
have this option under these circumstances.” Id. at 72,571.

Arizona contends that the January 2009 finding
constituted only a determination that Arizona failed to submit
a Section 309 SIP, and that the correct remedy for Arizona’s
asserted deficiency “was to impose a FIP supplying the
missing Section 309 elements, not to impose a FIP under
Section 308, as it did here.” But, as we explained earlier, see
supra, p. 11–12, Section 309 provides an alternative
mechanism for western states to comply with the CAA’s
visibility requirements for certain Class I areas. Such states
must include Section 308 components for other Class I areas,
and they remain subject to Section 308’s requirements if they
do not submit a Section 309 SIP. See 40 C.F.R. § 51.309(a),
(e), (g)(2). Arizona expressly acknowledged that its Section
309 SIP submission lacked certain requirements under
sections 51.309(d)(4)(viii) and (g). Because the State did not
submit an adequate Section 309 SIP, it did not submit an
adequate regional haze SIP. EPA was required to promulgate
 ARIZONA EX REL. DARWIN V. USEPA 47

a FIP to fill in the gap. See 42 U.S.C. §§ 7410(c)(1),
7491(b)(2)(A).

More practically, it is unlikely that a different outcome
would have resulted if EPA had provided Arizona with
additional time to correct its Section 308 SIP. EPA had
expressed a number of its concerns to Arizona after reviewing
its proposed SIP in late 2010. EPA also identified the basis
for its proposed partial disapproval in its July 2012 proposed
rule. Arizona made no effort to correct its SIP in light of
these comments. There is no reason to think it would have
done so after the Final Rule disapproving the SIP issued
either.

In sum, EPA properly promulgated its FIP in the same
rule as its partial disapproval of the SIP. Further, as EPA
stated in the Final Rule, the State remains free to, at any time,
“submit a revised SIP to replace the FIP.” 77 Fed. Reg. at
72,571.

IV. CONCLUSION

For the reasons set forth above, we deny Arizona’s and
SRP’s petitions for review of EPA’s Final Rule as to EPA’s
disapproval of Arizona’s SIP. Our ultimate review of EPA’s
FIP, however, must await EPA’s final action on its proposal
to revise the FIP in specific respects. As noted, EPA has
stated that it expects to finalize the revised FIP by March
2016. Accordingly we stay these proceedings as to
evaluation of the FIP’s technical feasibility until the
administrative process is complete.

PETITIONS DENIED.
48 ARIZONA EX REL. DARWIN V. USEPA

GLOSSARY OF ACRONYMS

AFUDC – Allowance for Funds Utilized During Construction

APA – Administrative Procedure Act

CAA – Clean Air Act

BACT – Best Available Control Technology

BART – Best Available Retrofit Technology

EPA – Environmental Protection Agency

FIP – Federal Implementation Plan

IPM – Integrated Planning Model

LNB – Low NOX Burner

NOX – Nitrogen Oxides

OAQPS – Office of Air Quality Planning and Standards

OFA – Overfire Air

PSD – Prevention of Significant Deterioration

SCR – Selective Catalytic Reduction

SIP – State Implementation Plan

SO2 – Sulfur Dioxide
 ARIZONA EX REL. DARWIN V. USEPA 49

SRP – Salt River Project Agricultural Improvement and
Power District